mongo-ruby-driver/lib/mongo/repl_set_connection.rb

# encoding: UTF-8

# --
# Copyright (C) 2008-2011 10gen Inc.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# ++

module Mongo

  # Instantiates and manages connections to a MongoDB replica set.
  class ReplSetConnection < Connection
    attr_reader :nodes, :secondaries, :arbiters, :secondary_pools,
      :replica_set_name, :read_pool, :seeds

    # Create a connection to a MongoDB replica set.
    #
    # Once connected to a replica set, you can find out which nodes are primary, secondary, and
    # arbiters with the corresponding accessors: Connection#primary, Connection#secondaries, and
    # Connection#arbiters. This is useful if your application needs to connect manually to nodes other
    # than the primary.
    #
    # @param [Array] args A list of host-port pairs to be used as seed nodes followed by a
    #   hash containing any options. See the examples below for exactly how to use the constructor.
    #
    # @option options [String] :rs_name (nil) The name of the replica set to connect to. You
    #   can use this option to verify that you're connecting to the right replica set.
    # @option options [Boolean, Hash] :safe (false) Set the default safe-mode options
    #   propogated to DB objects instantiated off of this Connection. This
    #   default can be overridden upon instantiation of any DB by explicity setting a :safe value
    #   on initialization.
    # @option options [Boolean] :read_secondary(false) If true, a random secondary node will be chosen,
    #   and all reads will be directed to that node.
    # @option options [Logger, #debug] :logger (nil) Logger instance to receive driver operation log.
    # @option options [Integer] :pool_size (1) The maximum number of socket connections allowed per
    #   connection pool. Note: this setting is relevant only for multi-threaded applications.
    # @option options [Float] :pool_timeout (5.0) When all of the connections a pool are checked out,
    #   this is the number of seconds to wait for a new connection to be released before throwing an exception.
    #   Note: this setting is relevant only for multi-threaded applications.
    # @option opts [Float] :op_timeout (nil) The number of seconds to wait for a read operation to time out.
    #   Disabled by default.
    # @option opts [Float] :connect_timeout (nil) The number of seconds to wait before timing out a
    #   connection attempt.
    # @option opts [Boolean] :auto_refresh (false) Set this to true to enable a background thread that
    #   periodically updates the state of the connection. If, for example, you initially connect while a secondary
    #   is down, :auto_refresh will reconnect to that secondary behind the scenes to
    #   prevent you from having to reconnect manually.
    # @option opts [Integer] :refresh_interval (90) If :auto_refresh is enabled, this is the number of seconds
    #   that the background thread will sleep between calls to check the replica set's state.
    #
    # @example Connect to a replica set and provide two seed nodes. Note that the number of seed nodes does
    #   not have to be equal to the number of replica set members. The purpose of seed nodes is to permit
    #   the driver to find at least one replica set member even if a member is down.
    #   ReplSetConnection.new(['localhost', 30000], ['localhost', 30001])
    #
    # @example Connect to a replica set providing two seed nodes and ensuring a connection to the replica set named 'prod':
    #   ReplSetConnection.new(['localhost', 30000], ['localhost', 30001], :rs_name => 'prod')
    #
    # @example Connect to a replica set providing two seed nodes and allowing reads from a secondary node:
    #   ReplSetConnection.new(['localhost', 30000], ['localhost', 30001], :read_secondary => true)
    #
    # @see http://api.mongodb.org/ruby/current/file.REPLICA_SETS.html Replica sets in Ruby
    #
    # @raise [ReplicaSetConnectionError] This is raised if a replica set name is specified and the
    #   driver fails to connect to a replica set with that name.
    def initialize(*args)
      if args.last.is_a?(Hash)
        opts = args.pop
      else
        opts = {}
      end

      unless args.length > 0
        raise MongoArgumentError, "A ReplSetConnection requires at least one seed node."
      end

      # The list of seed nodes
      @seeds = args

      # The members of the replica set, stored as instances of Mongo::Node.
      @nodes = []

      # Connection pool for primary node
      @primary      = nil
      @primary_pool = nil

      # Connection pools for each secondary node
      @secondaries = []
      @secondary_pools = []

      # The secondary pool to which we'll be sending reads.
      # This may be identical to the primary pool.
      @read_pool = nil

      # A list of arbiter addresses (for client information only)
      @arbiters = []

      # Refresh
      @auto_refresh = opts.fetch(:auto_refresh, true)
      @refresh_interval = opts[:refresh_interval] || 90

      # Are we allowing reads from secondaries?
      @read_secondary = opts.fetch(:read_secondary, false)

      # Lock around changes to the global config
      @connection_lock = Mutex.new
      @connected = false

      # Store the refresher thread
      @refresh_thread = nil

      # Replica set name
      if opts[:rs_name]
        warn ":rs_name option has been deprecated and will be removed in 2.0. " +
          "Please use :name instead."
        @replica_set_name = opts[:rs_name]
      else
        @replica_set_name = opts[:name]
      end

      setup(opts)
    end

    # Initiate a connection to the replica set.
    def connect
      @connection_lock.synchronize do
        return if @connected
        manager = PoolManager.new(self, @seeds)
        manager.connect

        update_config(manager)
        #BSON::BSON_CODER.update_max_bson_size(self)
        initiate_auto_refresh

        if @primary.nil? #TODO: in v2.0, we'll let this be optional and do a lazy connect.
          raise ConnectionFailure, "Failed to connect to primary node."
        else
          @connected = true
        end
      end
    end

    # Note: this method must be called from within
    # a locked @connection_lock
    def update_config(manager)
      @arbiters = manager.arbiters.nil? ? [] : manager.arbiters.dup
      @primary = manager.primary.nil? ? nil : manager.primary.dup
      @secondaries = manager.secondaries.dup

      @primary_pool = manager.primary_pool
      @read_pool    = manager.read_pool
      @secondary_pools = manager.secondary_pools
      @seeds = manager.seeds
      @manager = manager
      @hosts = manager.hosts
      @nodes = manager.nodes
    end

    # If ismaster doesn't match our current view
    # then create a new PoolManager, passing in our
    # existing view. It should be able to do the diff.
    # Then take out the connection lock and replace
    # our current values.
    def refresh
      return if !connected?

      if !Thread.current[:background]
        Thread.current[:background] = PoolManager.new(self, @seeds)
      end

      background_manager = Thread.current[:background]
      if update_struct = background_manager.update_required?(@hosts)
        @connection_lock.synchronize do
          background_manager.update(@manager, update_struct)
          update_config(background_manager)
        end
      end
    end

    def connected?
      @connected && !@connection_lock.locked?
    end

    # @deprecated
    def connecting?
      false
    end

    # The replica set primary's host name.
    #
    # @return [String]
    def host
      super
    end

    # The replica set primary's port.
    #
    # @return [Integer]
    def port
      super
    end

    # Determine whether we're reading from a primary node. If false,
    # this connection connects to a secondary node and @read_secondaries is true.
    #
    # @return [Boolean]
    def read_primary?
      @read_pool == @primary_pool
    end
    alias :primary? :read_primary?

    # Close the connection to the database.
    def close
      super

      @connected = false
      if @refresh_thread
        @refresh_thread.kill
        @refresh_thread = nil
      end

      if @nodes
        @nodes.each do |member|
          member.disconnect
        end
      end

      @nodes = []
      @read_pool = nil

      if @secondary_pools
        @secondary_pools.each do |pool|
          pool.close
        end
      end

      @secondaries     = []
      @secondary_pools = []
      @arbiters        = []
    end

    # If a ConnectionFailure is raised, this method will be called
    # to close the connection and reset connection values.
    # @deprecated
    def reset_connection
      close
      warn "ReplSetConnection#reset_connection is now deprecated. " +
        "Use ReplSetConnection#close instead."
    end

    # Is it okay to connect to a slave?
    #
    # @return [Boolean]
    def slave_ok?
      @read_secondary
    end

    def authenticate_pools
      super
      @secondary_pools.each do |pool|
        pool.authenticate_existing
      end
    end

    def logout_pools(db)
      super
      @secondary_pools.each do |pool|
        pool.logout_existing(db)
      end
    end

    private

    def initiate_auto_refresh
      return unless @auto_refresh
      return if @refresh_thread && @refresh_thread.alive?
      @refresh_thread = Thread.new do
        sleep(@refresh_interval)
        refresh
      end
    end

    # Checkout a socket for reading (i.e., a secondary node).
    # Note that @read_pool might point to the primary pool
    # if no read pool has been defined. That's okay; we don't
    # want to have to check for the existence of the @read_pool
    # because that introduces concurrency issues.
    def checkout_reader
      connect unless connected?

      if @read_secondary && @read_pool
        begin
          return @read_pool.checkout
        rescue NoMethodError
          warn "Read pool was not available."
        end
      end

      begin
        return @primary_pool.checkout
      rescue NoMethodError
        raise ConnectionFailure, "Not connected to any nodes."
      end
    end

    # Checkout a socket for writing (i.e., a primary node).
    def checkout_writer
      connect unless connected?

      if @primary_pool
        begin
          return @primary_pool.checkout
        rescue NoMethodError
        end
      end

      raise ConnectionFailure, "Failed to connect to primary node."
    end

    # Checkin a socket used for reading.
    def checkin_reader(socket)
      if @read_secondary
        @read_pool.checkin(socket)
      else
        checkin_writer(socket)
      end
    end

    # Checkin a socket used for writing.
    def checkin_writer(socket)
      if @primary_pool
        @primary_pool.checkin(socket)
      end
    end
  end
end